Testing users of the Repertoire client

When testing software that uses the Repertoire client for service discovery, the test suite needs to be able to inject test service discovery information and prevent attempts to contact a real service discovery service. rubin.repertoire provides a function, register_mock_discovery, to simplify this test setup.

Prerequisites

Since the Repertoire client uses HTTPX, register_mock_discovery relies on respx to mock the service discovery URL. The package being tested must therefore depend on respx in whatever dependency group it uses for tests that need service discovery mocking.

The Repertoire client does not itself depend on respx and expects the mock router to be passed into register_mock_discovery.

Mocking service discovery results

First, create the mocked service discovery results that you want the test to use. It’s often sufficient to create one set of mocked results that can be used by the entire test suite.

register_mock_discovery accepts these results in any of three formats:

• A Discovery object.

• A Python dict holding the serialized form of a Discovery object (the same as the JSON output produced by the Repertoire server’s /discovery route).

• A Path for a file containing the serialized form of a Discovery object in JSON format.

Then, call register_mock_discovery, passing in the respx_mock fixture object and the mocked discovery results. Normally, this is best done in a fixture defined in tests/conftest.py. For example:

import pytest
import respx
from pathlib import Path
from rubin.repertoire import Discovery, register_mock_discovery


@pytest.fixture(autouse=True)
def mock_discovery(
    respx_mock: respx.Router, monkeypatch: pytest.MonkeyPatch
) -> Discovery:
    monkeypatch.setenv("REPERTOIRE_BASE_URL", "https://example.com/repertoire")
    path = Path(__file__).parent / "data" / "discovery.json"
    return register_mock_discovery(respx_mock, path)

This fixture assumes tests/data/discovery.json contains the mock service discovery results in JSON format. It is set up as an autouse fixture that will affect all tests. If you need to vary service discovery results by test, you will need to do something more complex, such as multiple fixtures.

Note that this fixture uses pytest’s monkeypatch.setenv to set REPERTOIRE_BASE_URL to a base URL for the mocked Repertoire service. This is the recommended pattern, since this environment variable will also be used by the Repertoire client and ensures the client queries the mocked URL.

However, you can instead pass the base URL to register_mock_discovery as its third argument if you choose, in which case the environment variable will be ignored and does not need to be set.

This fixture returns the Discovery model. This allows tests to explicitly depend on the fixture and reference that model for expected URLs that show up in test results This can avoid accidental mismatches test URLs between the mock service discovery results and the tests.